Skip to Content

In Silico

Sanofi Bets on Schrödinger

Sanofi has signed an expanded deal with Schrödinger, the computational chemistry folks. Here’s something from the press release:

Schrödinger has made a number of key scientific breakthroughs in recent years in the areas of protein and ligand structure determination and potency prediction that promise to have a transformative impact on the discovery of drugs. The collaboration with Sanofi aims to deploy this and related technologies at a level that is unprecedented in the pharmaceutical industry.

.
I’d guess that this is going to involve the FEP calcuations that they’ve been talking about (blogged here). Schrödinger is also very strong in doing molecular dynamics simulations (for similar reasons), although, as with everything in this field, there’s room to argue about what that can do for you. So this will be very interesting to watch. I’m glad that Schrödinger’s technology is being given such a thorough real-world test, because that’s the only way to see what it can do.

37 comments on “Sanofi Bets on Schrödinger”

  1. Wavefunction says:

    Agreed that this is the best way to battle-test new computational technologies. Another story to watch out for is that of the Schrodinger affiliate Nimbus which just got $43 million for taking their (computationally discovered) ACC inhibitor lead into pre-clinical studies.

  2. Anonymous says:

    > I’m glad that Schrödinger’s technology is being given such a thorough real-world test, because that’s the only way to see what it can do.
    Do I sense a not-so-oblique reference to a certain SV startup that’s been making a lot of noise lately?

  3. steve says:

    The big problem is doing the animal experiments as you can never really tell if they are alive or dead. Sometimes they’re both.

  4. gippgig says:

    Schrodinger’s cat?

  5. Exglax says:

    $120 million?! Ha ha ha ha ha ha ha…(Sound of Schrodinger laughing all the way to the bank)
    Of course, with that kind of money this dumb experiment has to have backing from dumb senior management. I’d bet this can’t possibly be seen to fail, so a fair and thorough road-test is the last thing likely to happen here…

  6. Anonymous says:

    I like the idea of this experiment and give props to Sanofi for being the first to invest at such a significant level in computation. It might not work, but look how well the status quo in pharma is working now. Sure, Sanofi is putting a bet on something that has not been fully vetted in big pharma, but based on the FEP Pipeline post and everything else that Schrodinger has been publishing, it seems like one of the smarter bets that could be made.

  7. Chrispy says:

    I always thought the advantage of in silico chemistry was that it would encourage people to try things they might not otherwise try. A lot of the analogues of lead compounds made now are made because it is easy to make them (especially if management has decided to reward based upon the number of molecules).
    On the plus side for chemists, you really do need to test these things eventually, so someone will have to make them. And at least some of them are likely to be too difficult to outsource.

  8. Anon says:

    What’s happening here is very simple: Sanofi fired almost all of its in-house modelers and informatics guys over the last five years, so they have nobody left to contribute. Hence the collaboration with Schrodinger. It’s idiotic of course since with that amount of money they could have hired some world-class modelers and used many other algorithms and not just Schrodinger’s to advance their programs. One more example of how Big Pharma is sailing to hell in a handbasket lovingly crafted by its own senior management.

  9. Anonymous says:

    Agree with #8, except that the remaining modelers in Boston were laid off specifically to be replaced by Schroedinger. There still was internal capability, now to be replaced with contractors that will be only minimally embedded.

  10. Anonymous says:

    I love how this is sold as the solution. Over a long career this has only been useful for monday morning quarterbacking why a compound was active when it was expected to be not so. Thought it was sold as a result of the calculations. I have yet to see anything “discovered” from a computational model.

  11. seriously? says:

    sell sanofi stock right now

  12. Anonymous says:

    It’s just over the odds outsourcing to compensate for internal headcount reductions. They’ll never get anywhere near paying the milestones making the headlines.
    If you are lucky and have some good modellers and chemists working together and generating IP, I wouldn’t outsource it, or break things up to replace with a model of this sort…

  13. Drug hunter says:

    Derek has always been grounded when he’s said that the best modelers are strongly conversant with medchem and understand the limits of what they can and cannot model. Wavefunction is right, that this is an exciting battle-test of well-resourced new technology. Anonymous #12 is right that success is dependent on good modelers tightly integrated with good chemists.
    If drug hunting is your calling, and you have both strong org/med-chem training, exceptional 3D design skills, and are willing to work at a computer rather than a fume hood, Schrodinger is looking for more people (including training the next generation) to participate in these experiments with Sanofi, Nimbus, and beyond. Not to mention that the job security at a 25-year-old firm that’s the leader in its niche, with deep-pocketed backers, is different from both startup biotechs and larger pharma.

  14. MO says:

    @ 10
    “I have yet to see anything “discovered” from a computational model.”
    Really? See slides 18, 22
    http://cgen.com/images/pdf/2015_3%20Corporate%20presentation%20Final.pdf

  15. Anonymous says:

    @10: I do not think anyone here (or anywhere, for that matter) would say that drugs are “discovered” from computational models. Eventually, everything must be run through experiments (obviously). To be useful, the computational models do not need to be perfect — they simply need to help us make better decisions. We cannot all sit here pretending that we are making perfect decisions in the lab. We try to use intuition and past experiences to guide the drug design process, and it generally works better than random, but I see no reasons why computational models cannot help us make even better decisions, especially with the more legitimate methods that really try to model the physics right (e.g. free energy simulations). We all got burned on this line of thinking decades ago, and in hindsight it was clearly too early then, but the field has evolved. Algorithms are better, force fields are more accurate, and computers are a lot faster. This is clearly moving us toward models that should be more predictive. The big question is whether the best models are good enough now to make an impact, and improve the efficiency of the drug design process. Sanofi is betting on that. Maybe they are too early, or maybe we are too late. We will see. I am looking forward to tracking how this goes (if there is any way to actually do that).

  16. anon the II says:

    @13
    “including training the next generation” is code for “old farts need not apply”.

  17. Anonymous says:

    The sad thing is that we will never ever get to know about the how beneficil that deal was for Sanofi:
    If the collaboration turns out to be productive, they will not talk about it. If Schro fails to deliver…. well then neither Schro, nor Sanofi would talk about it.
    Also watch the details carefully: As far as I understood, Sanofi merely promized milestone payments if Schrodinger manages to deliver on topics like target validation, etc… How do you measure that? They clearly wont wait untill there really is a drug coming out of the first target computationally “validated” by Schro?
    No, no, as others have stated, the misery is that Sanofi sacked a lot of talent in their comp chem departments and now the buy in on a desperate deal.
    Sad state of the industry; not creative research management…
    Well done – once again – for the marketing guys at Schrodinger.

  18. Over50 says:

    Technophobia: “the fear or dislike of advanced technology… especially computers”.
    It’s depressing how many people commenting here are so cynical about “advanced technology”. How are we supposed to advance the field if we don’t do experiments (as Derek suggested)? Is everyone who is so aggressively opposed to this experiment really happy with the current state of affairs and convinced that there is no room for improvement? Maybe this particular experiment will fail, but why not try? Shouldn’t we all want it to succeed? Maybe the folks at Sanofi and Schrödinger actually care about saving lives and improving human health, just like the rest of us.

  19. Anonymous says:

    Sanofi was going to outsource a large portion of research no matter what. Schrodinger clearly is not the cause of the layoffs. If Schrodinger did not exist, then the layoffs would have still happened and some other comp chem organization would have gotten the contract. It looks like Sanofi chose the biggest/best company for the job. Based on the other articles in the Pipeline alone (plus all of the publications, collaborations, etc), we can see that Schrodinger is on the leading edge of the comp chem field, and bringing in the industry leaders in perfectly inline with the Sanofi strategy (at least per the previous CEO, Viehbacher).
    Viehbacher got the ball rolling years ago and it was clear that people at Sanofi were going to lose jobs. Viehbacher wanted to create an “ecosystem”, which had a mix (roughly 50/50) of internal and external research. There is a lot of momentum going that direction now and a little comp chem company like Schrodinger is clearly not going to be able to divert that ship. On the other hand, the new Sanofi CEO may have plans to outsource less, but that will take time.

  20. Over50 says:

    Technophobia: “the fear or dislike of advanced technology… especially computers”.
    It’s depressing how many people commenting here are so cynical about “advanced technology”. How are we supposed to advance the field if we don’t do experiments (as Derek suggested)? Is everyone who is so aggressively opposed to this experiment really happy with the current state of affairs and convinced that there is no room for improvement? Maybe this particular experiment will fail, but why not try? Shouldn’t we all want it to succeed? Maybe the folks at Sanofi and Schrödinger actually care about saving lives and improving human health, just like the rest of us.

  21. MedChemist_Anon says:

    Any self respecting modeler should seriously consider dropping Schrodinger software. Next it will be your company that falls for the sales pitch and decides it will be cheaper to out source your job to Schrodinger. From my experience Schrodinger lead the field in marketing but not method accuracy – so look at MOE, molsoft, OpenEye as alternatives.

  22. Anonymous says:

    @21: huh? Drop the software that has the best science and the company that is making the most serious effort to make an impact in drug discovery. I am not sure I understand your rationale.

  23. Anonymous says:

    @21: Publication record for the companies you mentioned:
    Schrodinger: 240 total publications (since 1994) – cited 7336 times
    Schrodinger: 180 publications since 2005 – cited 3754 times
    Molsoft: 113 total publications (since 2006) – cited 3112 times
    OpenEye: 59 total publications (since 2005) – cited 1857 times
    CCG: 25 total publications (since 2005) – cited 255 times
    So, it would appear that Schrodinger, Molsoft, and OpenEye are doing a good job of attempting to advance the field. OpenEye has on average more citations per publication as compared to the other companies. CCG is quite a bit lower than the others. Maybe none of this really matters, but it’s interesting nonetheless.

  24. a-nony-mouse says:

    @22: Try using Maestro sometime and you’ll see what @21 means. This is Schrodinger’s central software product, and it’s a hunk of (very slow) junk. Look at the scoring metrics for Glide, or SiteMap. Do they make sense to you? Good, maybe you can explain them to the rest of us. Try rotating your molecules just bit before using either program, and see if you get the same results. Ever try getting protonation states/with Epik? Or had your protein get prepped for modeling in some very dumb way by their prep tool?
    Sure you could probably say the same for some aspects of the software from all of the other vendors, but much of Schrodinger’s stuff stands out as being often poorly designed, buggy, and/or poorly (slowly) performing.

  25. Anonymous says:

    @24: I am anonymous #22. I have not used Maestro so I cannot respond to your specific questions, but I have a feeling that the Sanofi collaboration does not have much to do with rotating molecules in a graphical interface. Nobody is saying that the software is perfect/magic, but the science seems solid. Maybe part of the collaboration is aimed at using the software in force, with enough computational resources and expert users to really see how it works, and take a step beyond the usability of the graphical interface.
    @23: Thanks for the citation search. It can be helpful to add some numbers into the discussion. It would be particularly interesting to see how many patents include each of the different pieces of software, but that might be a topic for a different discussion.

  26. Anonymous says:

    @25: Searching USPTO (for example using http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=%2Fnetahtml%2FPTO%2Fsearch-adv.htm&r=0&f=S&l=50&d=PTXT&Refine=Refine+Search&Query=drug+AND+%22Chemical+Computing+Group%22+AND+software):
    drug AND “Chemical Computing Group” AND software: 161 patents
    drug AND Schrodinger AND software: 143 patents
    drug AND OpenEye AND software: 49 patents
    drug AND Molsoft AND software: 43 patents
    I believe CCG does not have any scientists doing basic research (or maybe they have one), but they have a nice user interface (MOE), so the above is very interesting. It appears that the companies that are doing the most science have fewer patents mentioning their software than the company that is focused almost exclusively on the user interface. Maybe this is the reason that it costs so much to discover drugs and why so many drug candidates that finally make it into the clinic fail. Maybe if more companies focused on using good science rather than software that is easy to use, pharma would be doing a better job of discovering new drugs.

  27. Anonymous says:

    Learned a bit more info from a person who knows the inside details. Most Sanofi modelers globally still have their jobs, and there are no plans for that to change. 3 modelers in Boston lost their jobs as part of a general restructuring that had nothing to do with Schrodinger. Their jobs were going to be outsourced whether or not Schrodinger was involved.

  28. Anonymous says:

    Anonymous, but CompChem PhD with 15+ years combined experience at software vendor and Pharma. As others have pointed out, Schrodinger is far from the only innovator in this space, and the potentially troublesome part for Sanofi is that Schrodinger is only innovative in a very small slice of what one needs to arrive at a candidate molecule management. Affinity is easy to test, so having modeling capabilities here is a nice to have, not a must have for most projects (not evening mentioning that affinity is only a small step toward potency or efficacy). Schrodinger is not a leader in many other computational areas, notably knowledge based methods, multi-parameter optimization, and chemical space exploration. If I’m under pressure to produce drug candidates, I would much rather be good at those 3 computational disciplines than the discipline Schrodinger claims to be best at. So, if Schrodinger will be telling Sanofi chemists what they need, it will be a very bad thing. If Sanofi chemists can stand their ground and get both the Schrodinger brass and rank-and-file to better support their process of Discovery, well then maybe there’s hope. Discovery organizations are always going to run the experiment…and if the experiment is cheap and high throughput, then computation is going to be relegated to rationalization, not prediction. Every medicinal chemist I know can formulate very logical next steps on their own using their well developed brains, and the nice thing is those brains aren’t fettered by file formatting scut work and crappy user interfaces.

  29. Anonymous says:

    Learned a bit more info from a person who knows the inside details. Most Sanofi modelers globally still have their jobs, and there are no plans for that to change. 3 modelers in Boston lost their jobs as part of a general restructuring that had nothing to do with Schrodinger. Their jobs were going to be outsourced whether or not Schrodinger was involved.

  30. Anonymous says:

    @28: Good thing that you have the whole drug discovery thing under control using knowledge-based approaches. I was under the impression that pharma discovery was in trouble, but you have assuaged my concerns. Let’s all go out and start analyzing old data to find new drugs. Yay!

  31. Empirist says:

    Two things:
    1) Software that helps a project team to do its work, especially if it helps it do its work successfully, is a win. Software that does not do this, good science or bad, is not so important. There is nary a bit of science in PyMOL, yet it has been staggeringly successful helping folks get their ideas across. Researchers do not want to “use software” – they have problems they would like to solve and are willing to look at computational tools.
    2) There’s 50+ years of chemical literature to draw from when one looks for new ideas. However many papers have been published by all the modelling companies combined, there are far more from the community at large. Biosym produced a large number of papers – is it even possible to purchase their software any more?
    Aside from the literature, ideas can come from customers and collaborators, many of whom work at the forefront of our business. Those in pharma have most of the data. much of which is proprietary. Academia, which has the people and the methodology, has been asking for collaborations since the 90’s if not earlier.

  32. thinker says:

    @21
    Drop Schrodinger software so Molsoft or OpenEye can take their job instead? What’s the difference?

  33. H2L says:

    Sorry for coming to this thread so late. Looks like it is not quite dead yet, so I will add a few words. I am cautiously optimistic about the contribution that modeling can provide in drug discovery. In my experiences, a good modeler with the right tools can almost always add value in a project. At the same time, we are not close to the point where modeling alone (without experiment) can get us anywhere. As such, I am always mystified by the extreme negative responses toward modeling that are seen in this blog (and in the med chem community in general). I was at a Gordon Med Chem GRC where the audience cheered when a presenter said that a docking calculation produced the wrong binding mode (for a GPCR homology model). There are some very, VERY hard things that we have to deal with in drug discovery. My understanding is that Schrodinger is not focusing on all of those (allostery, PPIs, positive/negative modulators, unknown off-target effects, cellular/animal/human outcomes, etc). Instead, they are focusing on a very specific problem — potency optimization. This is certainly not THE problem, but it is ONE important problem, and a better method to predict potency would be highly valuable (at least for me). I would love to make less dead compounds. If their FEP method can do that, then they have something that could dramatically save money and accelerate the hit-to-lead-to-candidate process (even if it is not perfect, which it is not), and Sanofi might be the first big pharma to see that. For the rest of us, we will have to wait and see…

  34. Anonymous says:

    @23: Just my two cents, but in my experience the science coming out of CCG and OpenEye has been a lot more upfront and self-aware than that coming from Schrodinger. I’d say that in assessing their own results, Schrodinger displays much more of the phenomenon of “believing their own hype”, regardless of how many papers they have.

  35. Anonymous says:

    @23:
    You’re cought by one big fallacy: The fact that Schrodinger invests A LOT of recourses in useless hype based publications (and even publications that have the one mere intent which is to cite other publications) does have nothing to do with “advancing the field”.
    The sheer number of publications and citations is a misleading number in the day and age of a publication industry that will publish anything regardless of scientific quality.
    Do not forget that Schrodinger is the only major comp chem company whose founders/associates are Professors at the same time and therefore profit from an extensive publication list. Others (Paul Labute at CCG or Ant Nicholls at Openeye) have no direct interest in publishing anything as long as it does not relate to directly explaining the theory of a new product. The latter they do as you will see if you search for their publications. On the contrary, read the papers describing Schrodingers new features. Or just ask them about details on their lately-so-hyped FEP method. They will not talk about the crucial and critical points but over-hype the thing elusively.

  36. Anonymous says:

    @35: Hey Ant (or is this Paul?), nice post.

Comments are closed.